Modelling memory decay after injuries using household survey data from Khartoum State, Sudan

Table 2 Alternative Poisson regression models for the relationship with time since injury, involving main effects only^a

Relationship with time since injury	Number of parameters in model	Akaike information criterion (AIC)	Likelihood ratio χ² compared with categorical model	Degrees of freedom (DF)	Estimated probability that an injury is still reported		Estimated time when reporting probability is 0.5 (months)^b
Relationship with time since injury	Number of parameters in model	Akaike information criterion (AIC)	Likelihood ratio χ² compared with categorical model	Degrees of freedom (DF)	In month 6	In month 11	Estimated time when reporting probability is 0.5 (months)^b
Categorical	21	5669.34			0.29	0.17
Log-cubic^c	14	5669.39	14.05	7	0.26	0.15	2.1
Log-quadratic^d	13	5667.43	14.09	8	0.26	0.15	2.2
Log-linear	12	5672.67	21.33	9	0.34	0.12	3.2
No effect of months	11	5841.57	192.23	10	1.00	1.00

^aData included for reported injuries in months 1–11 before interview. All models incorporate main effects of sex, age, urban/rural status, education and socioeconomic status, in addition to a random household effect
^bCalculated moving backwards in time, starting at the middle of the month preceding the interview
^cAlso includes quadratic and linear terms
^dAlso includes linear term

ISSN: 1471-2288